Structure Prediction Cautionary Tale

A given chemical compound, depending on the conditions, can adopt multiple crystal structures known as polymorphs. Predicting these polymorphs in order to control the crystallization process is a long-sought goal, and it is critical in drug development. But making such predictions remains a tough job. Adam J. Matzger and Saikat Roy of the University of Michigan have taken a close look at the accuracy of predicting polymorphs by revisiting the crystallization of 6-amino-2-phenylsulfonylimino-1,2-dihydropyridine (below), which was used as a test molecule in a crystal-structure-prediction blind test in 2001. Since then, a second polymorph has been found, and now, using a technique called polymer-induced hetero­nucleation, Matzger and Roy have found a third polymorph (Angew. Chem. Int. Ed., DOI: 10.1002/anie.200903285). Multiple calculation methods predict that form II is the most stable of the three forms, but, in reality, experimental measurements show that it’s actually the least stable. In addition, Matzger and Roy’s analysis of the 2001 test results found that nobody predicted the existence of form I, which is the most stable, or form III. Their observations suggest that extra caution is required with crystal structure prediction and that computational methodologies need to be improved.